Never Lose a Claude Code Prompt Again

The problem with Claude Code history

Claude Code keeps a conversation history, but it's scoped to individual sessions. Once you close a session or start a new one, finding that prompt from three days ago means scrolling through JSON files in ~/.claude/.

Maybe you wrote a system prompt that got Claude to refactor your auth module perfectly. You want to reuse it on another project. Good luck finding it.

Prompt Cellar fixes this. Every prompt is indexed by project, directory, branch, and timestamp. You search "auth refactor" and there it is.

How the hook works

Claude Code supports native hooks at ~/.claude/hooks/. Prompt Cellar uses this built-in mechanism. No wrappers, no shell aliases, no monkeypatching.

When you run pc setup, Prompt Cellar detects Claude Code on your system and writes a hook file. That hook fires on every prompt submission and captures:

• The full prompt text
• Working directory and git branch
• Timestamp and session ID
• Git remote and commit hash (if configured)

Everything is encrypted with AES-256-GCM before it leaves your machine. The hook adds roughly 50ms of latency. You won't notice it.

Session tracking that actually helps

Prompt Cellar groups your prompts by Claude Code session. You can see exactly what you asked during a particular debugging session or feature build.

Because it captures your working directory and git branch, you can filter by project. Working on three repos this week? Each one gets its own timeline.

The workflow is simple: write your prompt in Claude Code, use Claude Code as you normally do, and find that prompt later when you need it. Nothing changes about how you work.

Quick start

Three commands. Under a minute.

1. Install the CLI

2. Log in to your account

3. Run setup (detects Claude Code automatically)

After setup, open Claude Code and send a prompt. It shows up in your Prompt Cellar timeline within seconds.

No credit card required. Your first 100 prompts are free.

What you get